Letter on practically using z scores in Angola, by Saskia van der Kam

Dear Field Exchange,

I have the following comments to make about the article "The practical implications of using Z-scores: CONCERN's experience in Angola" (Field Exchange, Issue 1, May '97).

The article challenges the use of <80% median weight for height to determine admission for feeding programmes, advocating instead the use of minus 2 standard deviations weight for height cut-off point. While the standard deviation approach is more 'scientifically' correct I suspect that it has a price.

Comparison of the theoretical cut off points of standard deviations and percentage of the median shows that using the percentage of the median, younger children (up to 65 cm) are more likely to be classified as malnourished, whereas using the standard deviation method, children taller than 65 cm are more likely to be classified as being malnourished. The difference in cut-off points (weights of the children) between both methods grows gradually to 0.5 kg at 115 cm. Border-line malnourished tall children who would not have been admitted to feeding centres using the percentage of the median method will be admitted using the standard deviation method. Conversely, a number of borderline malnourished small children under 65 cm who would have been admitted using the percentage median will not be included using the standard deviation approach.

Although, more statistically correct, I believe that the standard deviation method is not justified because of differential vulnerability between these age groups. Science has shown that younger children are more vulnerable to malnutrition and its consequences. Thus, by including borderline malnourished older children at the expense of borderline malnourished younger children there is an automatic discrimination against a more vulnerable group. Because of this I cannot support a switch to standard deviation admission criteria as a priority.

The size of the group of children below 65 cm who would be excluded is unknown. Some people might argue that this group would not be vulnerable as traditionally a child under 65 cm is being exclusively breast-fed and is therefore nutritionally protected. However, this is not strictly true for emergency situations where there are often many low birth weight, sick, abandoned or orphaned babies as well as mothers who are not managing to breast feed well. As an indication of this, data from a therapeutic feeding centre in Kenema, Sierra Leone (September 1995), showed that 10% of admissions were children under one year old.

The article shows that the use of standard deviations resulted in 1.6 times more admissions in the CONCERN Angola programme. I suspect a lot of the extra admissions were older children, close to 110 cm or 115 cm as the biggest difference in cut-off point between the two methods occurs in this height range (0.5 kg at 115 cm). However, while the number of extra cases to be expected clearly depends on the age distribution in a population and on the number of borderline malnourished, I am surprised that it caused 60% more admissions.

This approach also raises an issue about overall food aid availability. As it is we often have to lower our admission criteria during emergencies as food runs out for a variety of reasons. Given that this situation is unlikely to change in the short-term, it seems to me unwise to expand feeding centre populations with more children who are only border- line malnourished.

The article correctly states that more stunted children are likely to be included using standard deviations. My question is whether this type of feeding programme is appropriate for stunted children. Surely, problems of chronic malnutrition are better dealt with in the public health sphere rather than through expanding feeding programme coverage.

The article also argues that the approach will allow for easier comparison between nutritional survey results and data from feeding centres. However, I feel that the advantage is limited as we now express survey results as percentage of the median (next to the standard deviation results). Also, methodological problems like the difficulty of comparing data from survey catchment areas with those of feeding centre catchment area are not solved by this statistical initiative.

While the article appreciates that the training of staff in concepts of standard deviation and its application will be a challenge, I really question whether in the highly resource constrained circumstances of an emergency, efforts required for such training would be far better expended in training on more vital concepts and operational practice.

The idea of using standard deviations for admission criteria for feeding centres is really revolutionary. But, I wonder whether overall, given the ensuing management problems and implications for food needs, it is an appropriate initiative. I suspect that the energy and resources needed to implement this type of approach would be better used in other activities like ensuring hospitalised children get proper food and improving food security and public health in general.

Yours etc,
Saskia Van der Kam,
Nutritionist, MSF Holland.